The AI Boom Is Creating a Financial System We Haven't Seen Before (Paul Kedrosky Explains)

Watch on YouTube ↗  |  August 28, 2026 at 15:45  |  45:15  |  Meb Faber Show
Speakers
Paul Kedrosky — Partner, SK Ventures

Summary

Paul Kedrosky discusses the massive scale and misunderstood exponential functions driving the current AI boom, comparing it to historical financial bubbles. He highlights how the data center buildout is increasingly debt-financed, how AI tokens are hyper-deflationary, and why an impending wave of mega IPOs could pressure current tech market winners.

  • AI token prices are deflating at 70-80% per year, requiring frontier AI companies to grow volumes by 400% just to stand still.
  • Over half of data center financing now comes from external debt rather than internal cash flows, increasing financial fragility.
  • Tech companies heavily investing in AI capex are taking on utility-like debt and maintenance obligations but remain wildly overvalued compared to orthodox utilities.
  • An upcoming wave of mega IPOs (estimated at over $5 trillion) will likely force institutional investors to sell their most liquid, best-performing tech stocks to free up capital.
  • Massive cash inflows into Taiwanese and Chinese chip manufacturers threaten to create a tsunami of semiconductor supply by 2028, potentially crashing prices.
  • AI is lowering the barriers to entry for chip design, threatening the moats of incumbent semiconductor companies and the perceived GPU duopoly.
  • Frontier AI models are converging in performance and experiencing diminishing returns, making multi-billion dollar training runs increasingly unjustifiable.
Ideas
Paul Kedrosky Partner, SK Ventures 18:44
Model convergence and deflation doom frontier AI.
Frontier AI models are experiencing rapidly diminishing returns in year-over-year performance gains and are converging in capabilities, making them virtually indistinguishable. Combined with the hyper-deflationary nature of AI tokens, these companies must grow unit volumes by 400% just to stand still, making their business models highly toxic and their massive multi-billion dollar training runs unjustifiable.
Paul Kedrosky Partner, SK Ventures 21:49
AI capex will rerate tech into utilities.
Technology companies heavily investing in AI data centers are increasingly relying on external debt financing rather than internal cash flows, saddling them with massive debt and ongoing maintenance capex. They are effectively turning into utility companies but are wildly overvalued compared to orthodox utilities, and will inevitably face a downward valuation rerate to match their new utility-like financial profiles.
Paul Kedrosky Partner, SK Ventures 24:53
Mega IPOs will force selling tech winners.
An unprecedented wave of upcoming mega IPOs (like SpaceX and Anthropic), estimated at over $5 trillion, will require institutional investors to free up massive amounts of capital. To fund these purchases, funds will structurally sell their most liquid, best-performing tech holdings that have overlapping factor exposures, putting significant downward pressure on current market winners.
Paul Kedrosky Partner, SK Ventures 30:09
Supply glut will crash semiconductor chip prices.
Unprecedented cash inflows into Taiwanese and Chinese chip manufacturers are setting up a massive tsunami of supply by early 2028. Because semiconductors are a highly capital-intensive boom-bust industry, this locked-in supply will inevitably crash prices as manufacturers dump inventory to cover fixed costs.
Paul Kedrosky Partner, SK Ventures 31:27
AI chip design destroys incumbent semiconductor moats.
The narrative of a permanent semiconductor supercycle and an unbreakable GPU duopoly is flawed because AI itself is now being used to design chips. This 'vibe chipping' drastically reduces design verification times and lowers traditional barriers to entry, meaning a flood of new, specialized chips will soon challenge incumbent market leaders and destroy their moats.
Up Next

This Meb Faber Show video, published August 28, 2026, features Paul Kedrosky discussing Frontier AI model companies, Technology companies tied to AI capex, Liquid Tech Winners, Taiwanese and Chinese chip manufacturers, SMH. 5 trade ideas extracted by AI with direction and confidence scoring.

Speakers: Paul Kedrosky  · Tickers: Frontier AI model companies, Technology companies tied to AI capex, Liquid Tech Winners, Taiwanese and Chinese chip manufacturers, SMH